Rescaled all ratings to range $[0, 1]$: higher = higher quality domain - independent fact-checkers (60) `fc` - NewsGuard (207) `ng`: www.newsguardtech.com - Media Bias/Fact Check (3216): www.mediabiasfactcheck.com - factual score `mbfc_fact` - bias score `mbfc_bias` (higher = higher quality, less bias) - `mbfc`: `mean(mbfc_fact, mbfc_bias)` - `mbfc_min`: `min(mbfc_fact, mbfc_bias)` - Ad Fontes Media (283): www.adfontesmedia.com - reliability (factual) score `afm_rely` - bias score `afm_bias` (higher = higher quality, less bias) - `afm`: `mean(afm_rely, afm_bias)` - `afm_min`: `min(afm_rely, afm_bias)` - Iffy unreliable sources index (471) `misinfome`: www.iffy.news - [google sheet](https://docs.google.com/spreadsheets/d/1ck1_FZC-97uDLIlvRJDTrGqBk0FuDe9yHkluROgpGS8/edit#gid=1144285784) - credible: mean ratings across urls to get mean domain rating - https://www.isthiscredible.com ![[Pasted image 20220302185310.png]] # Correlations ```python # quality01 cols = ["ng", "fc", "misinfome", "afm", "mbfc"] # fact + bias combined w = np.array([0.220, 0.280, 0.080, 0.220, 0.200]) ``` ## quality01 ![[Pasted image 20211101170907.png]] ## quality02 ```python # quality02 cols = ["ng", "fc", "misinfome", "afm", "mbfc"] # fact + bias combined w = np.array([0.050, 0.800, 0.040, 0.060, 0.050]) ``` ![[Pasted image 20211101171022.png]] ## quality03 ```python # quality03 cols = ["ng", "fc", "misinfome", "afm_min", "mbfc_min"] w = np.array([0.220, 0.280, 0.080, 0.220, 0.200]) ``` ![[Pasted image 20211101172359.png]] ## quality04 ```python # quality04 min(["ng", "fc", "mbfc_fact", "mbfc_bias", "afm_bias", "afm_rely"]) ``` ![[Pasted image 20211116230211.png]] ## quality05 ```python # quality05 # separate into 4 groups based on cutoffs: 0.25, 0.5, 0.75, 1.0 # for ["ng", "fc", "mbfc_fact", "afm_rely"], recode as such: # from: [0, 0.25), [0.25, 0.5), [0.5, 0.75), [0.75, 1.00] # to: [0.125, 0.375, 0.625, 0.875] mean(["ng", "fc", "mbfc_fact", "afm_rely"]) # of newly recoded columns ``` ![[Pasted image 20211118135122.png|00]] ## quality06 (consistent top/bottom quartiles) ```python # quality06 (uses bins in quality05) # Divided each measure into quartiles and checked for consistency across measures. # Then selected top/bottom quantiles. Gord selected/binarized bad/good domains. ``` ![[Pasted image 20211118170632.png|500]] ## quality07 (3332) ```python # quality07 cols = ["ng", "fc", "afm_rely", "mbfc_fact"] w = np.ones(len(cols)) / len(cols) ``` ![[Pasted image 20220302185907.png]] ## quality08 (3332) ```python # quality08 # all misinfome domains assigned 0, regardless of other ratings cols = ["ng", "fc", "afm_rely", "mbfc_fact", "mbfc_bias", "afm_bias", "misinfome"] w = np.ones(len(cols)) / len(cols) ``` ![[Pasted image 20220302191205.png]]